On Aug 22 2025, Arnd Bergmann wrote:
> On Thu, Aug 21, 2025, at 08:56, Benjamin Tissoires wrote:
> > On Jul 11 2025, Arnd Bergmann wrote:
> >>
> >> + break;
> >> + }
> >>
> >> + hid = dev->hid;
> >> + switch (cmd & ~IOCSIZE_MASK) {
> >
> > Jiri pinged me about this one, and I gave it a go with the existing
> > tests I have in selftests... and turns out that this changes the logic
> > of the ioctl processing.
> >
> > The removed block was in the default section of the switch/case
> > statement. Now it's added *after*, meaning that any ioctl that was
> > normally processed before are now caught in the default of the switch
> > statement below and return -ENOTTY.
> >
> > Running tools/testing/selftests/hid/hid_bpf showed that.
>
> Ah, of course, thanks for checking and describing the issue.
>
> Did you already come up with a fixed patch? I'm currently
> travelling and won't be able to send a v2 quickly, so if you
> have a version that works for you, let's use that instead.
>
I spent yesterday trying to find a way to make the old behavior looks
the same than the new, and got a rather big rewrite of your patch. I'll
send this out as a v2, keeping your From and SoB, but please shout if
you are unhappy with those 2 tags.
Cheers,
Benjamin
